Chenucol Engmeering Science, Vol. 46, No.

1 I, pp 2881-2894,
Printed in Great Britain.

1991.

ooo9-2509/91
53.00 + 0.00
CC 1991 Pergaman Press plc

THEORETICAL
AND EXPERIMENTAL
BUBBLE
FORMATION AT A SINGLE ORIFICE IN A
TWO-DIMENSIONAL
GAS-FLUIDIZED
BED

Department

J. A. M. KUIPERS,
W. PRINS and W. P. M. VAN SWAAIJ
Twente University of Technology, PO Box 217, 7500 AE
of Chemical Engineering,
Enschede, The Netherlands
(First received

17 September

1990; accepted in reused firm

7 March 1991)

Abstract-An
earlier developed,
first principles
hydrodynamic
model of gas-fluidized
beds has been
employed to study theoretically
bubble formation
at a single orifice in a two-dimensional
bed. For several
orifice discharge
rates, theoretically
predicted
bubble sizes, formation
times and shapes have been
Besides, the present experimental
compared
with experimental
data obtained from triggered photographs.
and theoretical
results were compared
with predictions
from two approximate
models reported
in the
literature
which are based on an idealized picture of the process of bubble formation.
The advanced
hydrodynamic
model appears
to predict the experimentally
observed diameters, formation times and
shapes of bubbles quite satisfactorily.
The observed and calculated
bubble diameters
fall between the
models. Both the experimental and theoretical results clearly indicate
predictions
from the two approximate
that a strong leakage of bubble gas into the surrounding
porous emulsion phase occurs, especially during
the initial stage of bubble formation.

1. INTRODUCTION

The formation
of gas bubbles is one of the most
characteristic
phenomena
of fluidized beds and it has
been recorded
over a considerable
span of time. Many
unique properties
of fluidized beds can be related
directly to the presence of bubbles and are dominated
by their behaviour. Therefore, accurate prediction
of
bubble characteristics
such as the size distribution,
the
bubble
rise velocity
distribution
and the bubble
frequency distribution
is practically important.
However, these distributions
depend on the initial bubble characteristics
at the gas distributor,
where the
bubbles are generated. Furthermore,
it is now widely
recognized that the gas-solids contacting efficiency of
fluidized bed chemical reactors is quite sensitive to the
bed hydrodynamics
just above the gas distributor
plate. Mass and heat transfer processes are seriously
affected by the mechanism
of bubble formation.
in
large scale gas-fluidized beds, normally, bubbles originate from discrete holes or other orfices in the gas
distributor
plate on which the bed rests. Bubble formation in gas-fluidized
beds at discrete orifices has
been studied experimentally
as well as theoretically
by
a number
of investigators.
Several
approximate
models, based on a strongly idealized picture of the
process of bubble formation,
have been presented in
the literature.

According
to the Davidson
and Schiiler model the
bubble volume Vb at detachment
and the corresponding time for bubble formation
t, (bubble detachment
time) are given respectively by
V, = 1.725(Co)0-h

g
(

>

and
fb

2
Q

(2)

where C, represents
the virtual mass coefficient of a
sphere, 9 the acceleration
of gravity and Q the constant gas flow rate through the orifice. The value of
the virtual mass coefficient C, is geometry dependent
(Davidson
and Harrison,
1963; Milne-Thomson,
lY60) and has been calculated for a number of relatively simple
configurations
of practical
interest
(Fig. 1). In the model adopted by Harrison and Leung
it has been assumed that gas leakage through
the
bubble surface into the emulsion
phase does not
occur. Nguyen and Leung (1972) injected air through
an orifice into an incipiently
fluidized two-dimensional bed of alumina particles. They correlated
the
observed
bubble volumes
with the gas flow rate
through the orifice and the frequency of bubble formation, nb, as

2. REVIEW AND DISCUSSION OF APPROXIMATE

THEORETICAL MODELS
Fluid&d
beds possess a number of fluid-like properties which led Harrison and Leung (1961) to adopt
the Davidson
and Schiiler (1960) model, originally
developed to describe spherical bubble formation at a
single orifice in an inviscid liquid, for the formation of
gas bubbles in fluidized beds under similar conditions.

indicating considerable
(47%) leakage of the injected
gas into the emulsion phase during the process of
bubble formation. Rowe et al. (1979) used X-ray cinephotography
to investigate the entry of gas from an
orifice into various fluidized powders. The technique
of X-ray observation
of bubbles in gas-Auidized beds,

J. A. M.

2882
,

orifice
disbibutor

KUIPERSet al.

(b)

orifice
plate

I-

emulsionphase

Fig. 1. Two geometrically different configurations for bubble formation at a single orifice in a gas-fluidized bed (threedimensional case). (a) C, = 1l/16, (b) C, = l/2.

developed by Rowe and co-workers,

has the advantage that the fluidized bed is not disturbed during the
experiment.
Analysis of their film showed that the
visible bubble flow could account only for approxjmately 50% of the gas injected into the bed. Obviously,
a considerable
gas leakage into the emulsion phase
occurs, just as in the experiments
of Nguyen and
Leung.
Yang et al., (1984) have studied bubble and jet
formation
in a 3 m diameter
semicylindrical
bed,
equipped with a similarly shaped 0.4 m diameter compound jet. In their large-scale cold-flow construction
with transparent
plexiglass pIates at the front (flat
side) and transparent
windows at the back (circumferential side), bubbles, and jets could be observed
directly against the plane wall of the bed. These
authors found that a substantial
amount of the gas
injected through the jet leaks from the bubble into the
emulsion phase, particularly
when the bed was operated below the incipient
fluidization
velocity. The
experimental
data were correlated in the same form as
eq. (3) with values of 0.303 or 0.118, depending on the
operating
conditions,
of the coefficient on the righthand side which indicates
a leakage of respectively
69.7% and 88.2%. Yang et al. (1984) developed
a
model to describe this phenomenon
by assuming that
a hemispherical
bubble is formed at the orifice where,
similar to the Zenz (1968) assumption,
bubble gas
leaks out through the bubbIe boundary at a superficial velocity equivalent
to the superficial minimum
fluidization velocity. The differential equation governing the rate of change of the bubble volume V, with
respect to the time f can then be expressed as:

where

Combination
of eqs (4) and (5) yields a differential
equation governing the rate of change of the bubble
diameter D, with respect to time which can be integrated with the initial condition,
D, = 0 at t = 0, to

give the following equation relating the time for bubble formation
t, and the corresponding
bubble diameter D,:

D h.mar=

Here 4, max represents the maximum bubble diameter, at which the total gas leakage through the
bubble boundary equals the total inflow through the
orifice (i.e. the jet). Equations
(6) and (7) and the
experimentally
determined bubble frequency nb = l/r,
provide the relations required to caIculate the bubble
diameter nh and the integral gas leakage 0, defined
by:
R = Qtb -

I(, =

Qt,- ;D:

Satisfactory
agreement
between the model predictions and the experimental
data was found. However,
a major drawback of the model proposed
by Yang
er al. (1984) is the necessary
input of the experimentally
observed
bubble frequency.
This experimental input is required because Yang et al. did not
employ a momentum
balance for the bubble. To
predict theoretically
both the (initial) bubble diameter
and the corresponding
bubble frequency,
the mass
balance
and momentum
balance
of the forming
bubble should be solved simultaneously.
This approach was followed by Caram and Hsu (1986) in a
theoretical model describing the formation of a spherical bubble in a fluidized bed.
Their equation of motion, describing the position of
the bubble centre S as a function of time, was identical
to the one employed in the Davidson
and Schiiler
model. Unlike the Davidson and Schiiler model adopted by Harrison
and Leung for spherical
bubble
formation
in Ruidized beds, the Caram and Hsu
model accounts for gas leakage into the emulsion
phase. Darcys law was used to obtain an expression
for the superlicial gas leakage velocity at the bubble
boundary.
Caram
and Hsu reported
satisfactory
agreement
of their model predictions
with limited
expcrimcntal
data obtained from the litcraturc.
The main objective of the present investigation
is to
compare experimentally
determined
sizes, of bubbles
formed at a single orifice in a two-dimensional
gasfluidized bed, with theoretical predictions
from a previously
developed
first principles
hydrodynamic
model of fluidized beds (Kuipers
et al., 1991). In
addition, the experimental
and theoretical results will
be compared with predictions from two approximate
models, both based on an idealized picture of the
process of bubble formation
in fluidized beds. Some
assumptions
of these approximate
models will be
discussed in relation to the results obtained from the
advanced hydrodynamic
model.

Bubble

formation

at a single orifice in a two-dimensional

The first approximate

model is in fact a modification of the Harrison
and Leung model (1961) and
describes circular bubble formation at a single orifice
in a two-dimensional
bed. As in the Harrison
and
Leung model it has been assumed that no gas leakage
through the bubble surface into the emulsion phase
occurs. In the second approximate
model it has been
assumed, following the Zenz (1968) assumption,
that
the superficial leakage velocity of the bubble gas at the
bubble
boundary is equivalent
to the superficial
minimum
fluidization
velocity. In terms of the approximate models the process of bubble formation
is
governed by the gas mass balance and the equation of
motion for the bubble, represented in generalized
form respectively by:

&iJ =Q -

Urn,4

where pe represents the emulsion phase density, S the

distance between the bubble centre and the orifice and
A, and V, the bubble surface and the bubble volume,
respectively. In the case of circular bubble formation
in a two-dimensional fluidized bed eqs (9) and (10) can
be written as (for the case of constant emulsion phase
density):

(12)
where 6, represents the bed thickness of the twodimensional bed. Equations (11) and (12) can be
solved, by specifying appropriate initial conditions, to

0.57 m

yield the
function
S = R,.
solution
must be

bed

2883

bubble radius R, and the bubble centre S as a

of time. Bubble detachment will occur when
Except for the case of unt, = 0, no analytical
of eqs (11) and (12) can be obtained and they
integrated numerically.
3. EXPERIMENTAL.

3.1. Equipment and experimental procedure

The experiments were carried out in a thin twodimensional
gas-fluidized bed shown schematically in
Fig. 2. Two-dimensional
fluidized beds have often
been used to study fluidization phenomena experimentally (Rowe, 1971) because of the possibility of
visual observation of the bed behaviour, for instance
by photography.
The experimental construction consists of a fluidized bed section built of 0.015 m thick
glass plates (internal bed dimensions:
width 0.57 m,
height 1.0 m and depth 0.015 m) and a gas distributor
section of 0.015 m-thick transparent plexiglass plates.
The fluidizing gas is introduced through a porous
plate (sintered stainless steel, average pore size 20 pm)
provided with a central rectangular pipe (internal
dimensions 15.0 mm x 15.0 mm). Humidified air was
used to reduce electrostatical charging of the bed.
Through the central rectangular pipe, covered with a
stainless steel wire mesh, secondary (humidified) air
could be injected independently.
Calibrated
thermal
mass flow controllers
were used to ensure constant
mass flow rates of both the primary and secondary
fluidizing air. Application of rapidly responding magnetic valves coupled to a micro-computer, allowed
accurate injection of secondary air quantities through
the central pipe. During the experiments the outflow
opening at the top of the fluidized-bed section was
covered with a stainless steel wire mesh to prevent
particle carry over from the bed and was exposed to
the atmosphere. In order to ensure a true two-dimensional behaviour of the bed in the entrance region

3- flange
l

gas-fluidized

/5izA

fluidized bed section (glass)

I- flange
-gas distributor

section (plexiglass)

- 3 mm glass beads

Fig. 2. Schematic representation of the two-dimensionalgas-fluidizedbed

J. A.

2884

M. KUIPERS et al

could be registered. Each photograph

shows a different stage during the bubble formation
process. A
measuring
grid at the front side of the two-dimensional bed was used to determine
the size of the
bubbles from the photographs.
Two alternative procedures were considered
for the evaluation
of the
equivalent bubble diameter D, from the photographs.

near the orifice, which is of particular importance

in
the present study, it was decided to use a rectangular
orifice pipe instead of a circular one.
The bed material consisted of spherical solid particles (narrow size range; average particle diameter d,
= 500 pm; microscopic
density pb = 2660 kg/m3). In
the experiments
solid particles
were added to the
incipiently fluidized bed until the bed level reached
the same height which was chosen for the simulations
(h,, = 0.5 m). The minimum fluidization
velocity u,~
was obtained
in the usual way by determining
the
point of intersection
of the bed pressure drop versus
fluidizing velocity curves for fixed bed and fluid bed
flow regime.
In a typical experiment
to study bubble formation,
initially both the primary
and secondary
air were
injected at minimum fluidization
velocity by two separate flow controllers while a third flow controller was
purging the air stream required to generate a bubble
at the central orifice. By an appropriate
micro-computer controlled switching of a carefully selected combination of magnetic valves, the secondary air stream
could be replaced by the latter one at the beginning of
each experiment.

(a) Calculation
of the equivalent
bubble
D, on basis of an elliptical bubble shape:

(13)

D, = J%%

where D, and D, represent respectively the measured

vertical and horizontal
bubble diameter.
(b) Calculation
of the equivalent
bubble diameter
D, from the measured bubble area A:
(14)
The reported average bubble diameters are based on
two or more bubbles for each time delay used in the
experiments.

3.2. Measurement of bubble properties

Photography
has been applied as a technique
to
measure bubble sizes during the process of bubble
formation
at the central orifice. A NIKON
F-301
camera was used together with an ILFORD
PANF
IS0400 film. With the aid of a micro-computer
a
pulse was generated
which triggered the camera to
take a photograph
of the bed while a flashlight was
generated
at the rear side of the bed. By applying
increasing
time delays between the generation
of a
pulse and the injection
of gas through
the central
orifice, the complete process of bubble formation

Table 1. Hydrodynamic

diameter

4. HYDRODYNAMIC

MODEL

Our previously
developed
theoretical
model of
gas-fluidized
beds (Kuipers et al., 1991) is based on a
two-fluid model approach
in which both phases are
considered
to be continuous
and fully interpenetrating. In fact, the equations employed in this theoretical model can be seen as a generalization
of the
Navier-Stokes
equations
for two interacting
continua.
Two sets of conservation
equations are used, governing the balance of mass, momentum
and thermal
energy in each phase. Table 1 shows the mass and

model equations in vector notation

Continuity equations
Fluid phase

ah4
-

+ (V*Ep,U) = 0

at

(Tl-1)

Solid phase
act1 - E)YSI

+ [v*(l
- E)P,V]

at

(Tl-2)

0.

Momentum equations
Fluid phase
4EPA

at

+ (v*&p,UU)= -

2
Evp

p(U

V) +

i+(v*U)~

II
(U-3)

Solid phase

au1 ~ E)P,VI
at

+ [V*(l - E)&VV]= - (1 - E)VP + fi(u - v) +

v (I - E) - &(V.iI
i.[

+ (v.(I - E){P,[(vv) + (v+]})

- G(E)VE+ (I-

II

E)p,g (Tl-4)

Bubble

formation at a single orifice in a two-dimensional

momentum
conservations
equations for both phases
in vector form. In the present study bubble formation
in a cold-flow two-dimensional
gas-fluidized
bed-will
be studied and because of the anticipated
small heat
effects, the solution of the thermal energy equations is
not considered
here. Due to the mathematical
complexity of the equations of change, a numerical solution method has been used (Kuipers et al., 1991).
The numerical technique has been embodied in an
unsteady two-dimensional
computer code written in
VAX-PASCAL.
The computer
model calculates the
porosity,
the pressure, the fluid phase temperature
and the solid phase temperature
and the velocity fields
of both phases in two-dimensional
Cartesian or (axisymmetrical)
cylindrical coordinates.
These variables
constitute the so called primary or basic variables.
For closure of the set of balance equations specification of the constitutive
relations is required which
implies specification
of all other variables in terms of
the basic variables. Incorporation
of these constitutive equations introduces
the necessary empirical information.
4.1. Constitutive equations
Fluid phase density p/ and solid phase density pS:
The fluid phase density is related to the pressure and
Buid phase temperature
by the ideal gas law:

Mr

pr
=mpFor the solid phase microscopic incompressibility
was
assumed. Accordingly a specified constant density p.,.
was taken:
(16)

P, = P&O.

Interphase
momentum
transfer coefficient 0: For
porosities E -c 0.8 the interphase
momentum
transfer
coefficient has been obtained
from the well-known
vectorial Ergun equation (see Radestock and Jeschar,
1971):

p=

150(l
-7 - 8
E

Pr
(U,)

+ 1.75(1 - E) (@rip) Ill-

gas-fluidized bed

where
Re

(184

c
d

U8W
to the Reynolds

;[IP +0.15(Rep)0.687],Re,

< 1030

0.44,

P-

Re,

EPf

lu - vld,
p,

(184

In eq. (18a), f(e) accounts for the presence of other

particles in the fluid and corrects the drag coefficient
for a single isolated particle.
Fluid phase shear viscosity hJ and solid phase shear
viscosity ps: The present model incorporates,
as evident from inspection of the momentum equations given
in Table 1, Newtonian
behaviour
with zero bulk
viscosities for both phases. Consequently
only specification of the shear viscosities is required here. For
the fluid phase shear viscosity pf the corresponding
microscopic
shear viscosity of the fluidizing medium
[air at approximately
under atmospheric
conditions
(p = 101.3 kPa, T = 293 K)] has been used in the
present

study:
p, = 2.0 x 10e5 Pas.

(19al

Direct rheological measurements

by Schiigerl et al.
(1961) indicate
that fluidiied
suspensions
exhibit
approximate
Newtonian behaviour at low shear rates,
and that the apparent bed viscosity, depending on the
bed conditions
(fluidization
velocity), particle type
and particle diameter, ranges typically between 0.5
and 1.2 Pas (i.e. 500 to 1200 times the viscosity of
water at room temperature).
Grace (1970) employed
an indirect method (bubble shape measurements)
to
estimate the apparent bed viscosity of fluidized suspensions
and found excellent
agreement
with the
results obtained by Schiigerl et al. (1961). Apparent
bed viscosities, for spherical ballotini with a narrow
particle-size
distribution,
have been
determined
experimentally
by Schiigerl et al. (1961) and empirically by Grace (1970). Their results are presented in
Table 2. The apparent bed viscosity of fluidized suspensions is rather high in comparison
with the viscosity of ordinary liquids. In the present study the bed
material consisted of mono-sized spherical glass beads
(ballotini) with particle diameter d, = 500 pm. Based
on the results obtained by Schiigerl and Grace the
following estimate for the solid phase shear viscosity

Table 2. Apparent

whereas
for porosities
E > 0.8 the
interphase
momentum
transfer coefficient has been derived from
the correlation
of Wen and Yu (1966):

The drag coefficient

Cd is related
number Re, by (Rowe, 1961):

VI
(17)

f(E) = &-2.65.

2885

Vc)

bed viscosities for narrow

spherical ballotini

Particle diameter
fuml
550
460
220
170
140
120
82
60

Apparent
Results
of S&tiger1

size range

bed viscosity

Pas

Results
of Grace
Pas
0.95

1.20
0.90
0.80
0.80
0.80
0.60
0.40

0.85
0.75
0.80
0.85
0.90
0.70

TMeasurements in a rotating cylinder viscometer.

1000

$Deduced

from bubble

shape measurements.

J. A. M. KUIPERS et al.

2886
/L~

Table 3. Data for the numerical simulations

has been made:

,us =

1.0 Pas.

(19h)

in the present study, the solid phase shear viscosity

p* has been taken to be constant as a first approximation. Thus, it has been assumed that & is independent
of the porosity E.
Solid phase elastic modulus G(E): As discussed in
the previous paper (Kuipers et al., 1991) the solid
phase elastic modulus G(E) is important
from both a
physical and a numerical
viewpoint
and has been
incorporated
in the model according
to the following
generalized form:
G(E) = - G,{exp[c(&*

-&)I}

Minimum Ruidization porosity

Minimum fluidization velocity
Orifice velocity

0.402
0.250 m/s
5.0, 10.0, 15.0 m/s

Particle diameter
Particle density

5.00x IOe4m
2660 kg/m3

Orifice diameter
Bed width
Initial bed height

1.50 X to-2m
0.51 I11
0.50 m

Initial freeboard pressure

101,325.o

X-grid size
Y-grid size

7.50 x lO-m

Time step

2.50 x 10-4s

1.25 x lo-rn

(20)

where G, represents the the normalizing units factor, c

the compaction
modulus and E* the compaction
gas
phase volume fraction. To prevent unacceptable
bed
compaction
G, has been taken as 1 Pa, c = 100 and
&* = 0.45.
4.2. Numerical simulation
Figure 3 shows the initial and boundary conditions,
used for the present numerical simulation
of bubble
formation
in a cold-flow
two-dimensional
gas-fluidized bed. The corresponding
numerical data are listed
in Table 3. The left- and right-side walls of the bed
were modelled as impermeable,
no slip rigid walls for
both phases. The gas distributor
at the bottom of the
bed was conceived as a prescribed
fluid phase influx
wall and an impermeable
no slip rigid wall for the
solid phase. At the top of the bed a continuous

outflow wall and an impermeable

no slip rigid wall
were assumed for the fluid phase and the solid phase,
respectively. Continuous
outflow implies that the fluid leaves the bed at its own chosen rate with minimal
upstream flow disturbance.
In accordance
with the
experimental
situation
the minimum
fluidization
condition was prescribed as the initial condition for
the numerical calculations.
It implies a force balance
between the buoyant weight of the suspended
solid
particles and the frictional forces exerted by the fluidizing gas; as a consequence
of this hydrodynamic
equilibrium
no net particle movement
occurs. The
initial bed height h,, was identical to the experimental minimum fluidization
bed height. As Fig. 3
shows, a freeboard of the same size as the initial bed
height h,, was provided to allow for bed expansion.

Pa

initial conditions
E=l
p=pO
ux= 0 y= Umf
vx=v

E=&

mf

=o
x

=o

Cmf
y

5nf

=V=O
7.
Y

P =P,,+

no slip rigid wall for both phases

prescribed

ta

continuative

Fig. 3. The initial and boundary

fluid phase influx wall, noslip

fluid phase outflow

(I-~,~)(p,-p~,,)g(h~i

rigid wallfor

solid phase

wall, no slip rigid wall for solid phase

conditions for the numerical simulation

two-dimensional
gas-fluidized
bed.

of bubble formation

in a cold-flow

Bubble

Fig. 4. Photographically

formation

al a single orifice in a two-dimensional

gas-fluidized

observed and theoretically

calculated bubble growth
dimensional
gas-Auidized bed (u, = 10 m/s).

bed

at a single orifice in a two-

2887

2888

J. A. M.

KUIPERS

et al.

At zero time the gas velocity injected through the

central orifice was increased
instantaneously
from
minimum
fluidization
velocity u,,,~ to the required
orifice velocity u,. Only three orifice velocities (i.e. 5,
10 and 15 m/s) were considered in this study. To save
computer time, symmetry about the ccntre line of the
bed (x = 0) was assumed which can be justified on the
basis of the symmetrical
initial and boundary
conditions. In the actual calculations
only the region to the
right of the bed centre line was considered
(required
number of computational
cells: 38 x 80 = 3040) with a
fictitious impermeable
free slip rigid wall for both
phases at x = 0 (i.e. the centre line of the bed).
5. RESULTS

Theoretically
calculated
and experimentally
observed bubble formation
for three different orifice
velocities, namely u, = 5, 10 and 15 m/s will he discussed. In addition, the theoretical
and experimental
results will be compared
with two approximate
models reported
in the literature.
Detailed results
however
are presented
only for the case of U,
= lOm/s.
5.1. Detailed resultsfor u, = 10 m/s
Figure 4 shows a comparison
of photographs
with
corresponding
density
plots for U, = lOm/s.
The
measuring grid visible in the photographs
is the same
as the computational
grid used for the numerical
simulations
(6x = 0.0075 m, 6y = 0.0125 m). The
density plots, shown in Fig. 4, have been obtained
from the calculated
instantaneous
solidity distributions according
to the procedure
described
in the
earlier paper (Kuipers et aE., 1991). In these plots the
local dot density is a measure for the local instantaneous solidity (1 ~ c).
A reasonable
similarity between the photographs
and the density plots can be observed. However, the
agreement between the photographs
and the density
plots is not perfect, especially near the roof of the
bubble where some discrepancies
between experiment
and theory can be observed. The photographs
show
subtle deviations from symmetry about the heart line
of the fluidized bed and aIso show a much sharper
porosity
transition
near the bubble roof. A better
resolution near the roof could probably be obtained
by using a much finer computational
grid; however, to
keep the computer
time reasonable,
this was not
attempted
in the present study. When judging the
theoretical
results depicted in Fig. 4 it must first be
borne in mind that the present model contains no
adjustable
parameters,
and second that, unlike previous theoretical
approaches
(Harrison
and Leung,
1961; Zenz, 1968; Caram and Hsu, 1986) no specific
assumptions
concerning
the mechanism
of bubble
formation have been made. Further inspection of Fig.
4 shows that, especially
during the final stage of
bubble formation,
both the theoretical
and experimental bubbles have a practically circular shape. A
comparison
of the theoretical
and experimental
results for the other orifice velocities (i.e. 5 and 15 m/s)

CM)
I
I

-0.10

-0.05

0.00

0.05

0.10

Fig. 5. Theoretically calculated porosity contours near a

detached bubble at t = 0.200 s (u, = 10.0 m/s).

yields quite similar results. Figure 5 shows a number

of porosity contours near the detached bubble at time
t = 0.2 s. In agreement
with the qualitative
results
from Fig. 4, it can be seen that very sharp porosity
gradients exist near the bubble base; near the bubble
roof these gradients are considerably
weaker.
Figure 6 allows a quantitative
comparison
of the
theoretically
calculated
and experimentally
determined bubble sizes as a function of time, for the case
of u, = 10 m/s. To determine bubble diameters from
the numerically calculated porosity distributions,
the
bubble contour was defined as a void fraction of 0.85
in the present study. It was found that this particular
choice defines the bubble boundary as a contour with
very strong porosity
gradients
(especially near the
bubble base) which is consistent
with experimental
observations.
Table 4 shows the sensitivity of the
theoretically
calculated horizontal, vertical and equivalent bubble diameters with respect to the adopted
bubble definition. The equivalent bubble diameter D,
was obtained from the numerical integration
of the
area A for which E > 0.85 and the definition according
to eq. (21).

(21)
To determine the bubble size as a function of time,
during the process of bubble growth, each 10 ms
triggered photographs
of the bed were taken. For the
interpretation
of the experiments
it was found necessary to correct for the time delay (approximately
20 ms) caused by the magnetic valves. As mentioned
earlier two procedures
for the evaluation of the equivalent bubble diameter D, from the photographs
were
considered
[see eqs (13) and (14)]. During the final
stage of bubble formation
the deviation
from the

Bubble formation at a single orifice in a two-dimensional

2889

gas-fIuidized bed

+ Experimental data

Fig. 6. Theoretically calculated and experimentally observed bubble growth at a single orifice in a twodimensional gas-fluidized bed (u, = 10 m/s). The predictions according to two approximate literature
models, which are based on an idealized picture of the bubble formation process, are also shown.

Table 4. Sensitivity of the calculated horizontal, vertical

and equivalent bubble diameters with respect to the adopted
bubble definition
Bubble definition
0.80

E > 0.85

E > 0.90

0.183 m
0.201 In
0.181 m

0.172 m
0.182 m
0.164 m

0.156 m
0.155 m
0.143 m

E >

D,
D
0,

been neglected.
For the case that the superficial
leakage velocity at the bubble boundary
equals the
superficial minimum fluidization
velocity (second approximate model), the bubble diameter as a function
of time has been obtained from integration
of the gas
mass balance eq. (11) to yield

hnf t
D?

+D,=__
Dbm

for

t < tb

(24)

where
circular bubble shape was always very small (D, x D,,)
and both procedures
yielded essentially
the same
results. However, due to the flat bottom of the bubble
during the initial stage of bubble formation,
the
assumption
of an elliptical bubble shape resulted in a
consequent

underprediction

of the more reliable

equi-

valent

bubble diameter obtained from the measured

bubble area A and eq. (14). In Fig. 6 the experimental
equivalent bubble diameter D, was obtained from the
bubble
measured
bubble
area A. The calculated
growth according
to the two approximate
models
presented in Section 2 is also shown in Fig. 6. For the
case of u,~ = 0, which implies no gas leakage through
the bubble boundary, the bubble diameter as a function of time has been calculated from
D, = 2
where the time for bubble

for
formation

t, = [ (F)($$)].

t-ctb

(22)

tb is given by
(23)

In eq. (23) C,, represents the virtual mass coefficient

of a cylinder which equals 1.0 for a moving cylinder in
an unbounded
fluid medium (Milne-Thomson,
1960).
To calculate the bubble formation
according
to the
approximate
models, C, has been taken to be 1.0. This
particular
choice implies that the effect of the gas
distributor
plate, which in fact restricts the movement
of the emulsion phase during the bubble growth, has

%4

Db = -.

Equation
(25) defines
the maximum
attainable
bubble diameter achieved when the total gas leakage
through the bubble boundary equals the total inflow
through
the orifice. However,
at the condition
of
bubble detachment
(S = Rb) the bubble diameter is
always smaller than the asymptotic
diameter calculated according
to eq. (25) which implies that the
situation of maximum leakage does not occur. The
time for bubble formation
t, has been obtained
by
simultanteous
numerical
integration
of eqs (11) and
(12) with a fourth order Runge-Kutta
method, S = R,
being the imposed condition
for bubble detachment.
However, it must be noted that the computed bubble
growth curves from the approximate
models, shown
in Fig. 6, have been obtained from the solution of the
respective gas mass balances only. In fact, according
to the approximate
models, bubble growth terminates
at t = tb (i.e. at t = 0.20s and t = 0.15 s for the model
with and without gas leakage, respectively) due to the
bubble detachment
from the orifice and the process of
bubble formation
repeats itself. Inspection
of Fig. 6
shows that the agreement
between the predictions
from the advanced
hydrodynamic
model and the
experimental
data is satisfactory.
The hydrodynamic
model tends to predict somewhat larger bubbles than
the experimentally
observed
ones which may be
caused by the wall effects in the two-dimensional
bed.
These wall effects (due to the presence of the front and

J. A. M.

2890

KUIPERS

back wall in the experimental

bed) are not included in
the present two-dimensional
model. To investigate
this effect theoretically,
an extension
of the twodimensional
model to a three-dimensional
model
would be required.
Figure 6 also shows that, initially, both approximate models
predict
essentially
the same bubble
diameters.
At the same time both models initially
overpredict
the experimentally
observed bubble diameters. It can be shown, by making a Taylor expansion on the left-hand
side of eq. (24) and subsequent rearrangement
of the resulting expression,
that eq. (24) reduces to eq. (22) provided that

For short times, condition (26) is fulfilled and thus

both approximate
models
should
indeed predict
essentially the same bubble diameters during the initial stage of bubble formation.
With respect to the
second observation
it should be realized that the
modified Harrison and Leung model overpredicts
the
experimentally
observed bubble diameters during the
entire process of bubble formation, because it assumes
zero gas leakage through the bubble boundary. During the final stage of bubble formation, the predicted
bubble sizes fall below the experimentally
observed
bubble sizes in case bubble gas leaks through the
bubble boundary
at a rate equal to the superficial
minimum fluidization velocity. This phenomenon
can
be explained by the increasing leakage rate of bubble
gas through the bubble boundary due to the increasing exchange area with the emulsion phase A,. From
the experimental
data shown in Fig. 6 it can be
deduced that initially strong leakage of the bubble gas
into the emulsion
phase occurs. Through
the introduction
of the integral leakage fraction Y(t), defined

as

Y(t) = 1 -

b() _

Yo(Q

for a two-dimensional
the integral

gas leakage

geometry

from a growing

bubble,

(27)
with

el

al.

an actual bubble volume Va(t) and a corresponding

(equivalent) bubble diameter D&), into the surrounding emulsion phase can be quantified. In eq. (27) &(t)
and D,(t) represent the bubble volume and the corresponding bubble diameter respectively, calculated according to the modified Harrison and Leung model
which assumes zero gas leakage. Figure 7 shows a
comparison
of the integral leakage fraction as a function of time, calculated from the experimentally
determined bubble growth, with the corresponding
theoretical predictions
from the numerical and approximate models. Figure 7 shows the integral leakage fraction Y(t) for t < r, only. In each case the final data
point indicates
that, at the corresponding
time t,
bubble detachment
has occurred.
Both the experimental data and the predictions
from the advanced
hydrodynamic
model indicate that the leakage rate
during bubble growth decreases.
This (decreasing)
quantity
can be expressed
as the product
of the
(bubble surface) averaged superficial leakage velocity
and the increasing
(due to bubble growth) bubble
surface. Obviously, the strongly decreasing superficial
leakage velocity overcompensates
for the effect of the
increasing bubble surface. Both approximate
models
do not account for this phenomenon
properly. Not
even a qualitative
agreement
with the experimental
of a
data can be observed. Obviously, the assumption
uniform
exchange
velocity throughout
the entire
bubble
boundary,
as made in these models,
is
incorrect.
Figure 8(b) shows the numerically calculated mass
flux profile for the gas phase at the boundary
of a
circular bubble at time t = 0.2 s, that is just after
bubble detachment.
This mans flux has been plotted
versus the angle with the heart line of the fluidized
bed. Figure 8(a) is meant to elucidate the way in which
Fig. 8(b) has been constructed.
It is a diagram determining the position of the circular bubble with respect
to the position of the calculated
bubble. S and rb
denote respectively the bubble centre and the bubble
radius. First note that the numerically
calculated
bubble contour F = 0.85 at time f = 0.2 s can be represented very well by a circle. For a circular bubble

+ Experimental data

Q = uadz

Fig. 7.

Integral

leakage fraction Y(t), calculated from eq. (27), as a function of time for bubble growth at a
single orifice in a two-dimensional gas-fluidized bed (u, = 10 m/s).

Bubble

formation at a single orifice in a two-dimcnsiuual

o.30 Figure 8a

gas-fluidized

bed

2891

ii = (sin(Q), cos(@))

(4
0.25

t
Y

0.20
0.15
0.10
0.05
0.00
-0.20 -0.15 -0.10 -0.05 -0.00 0.05 0.10
x +
(m)

0.00

0.20

0.40

0.60

0.15

0.80

0.20

1.00

Fig. 8. Calculated
gas phase mass flux profile at the boundary of a circular bubble with centre S and radius rb approxmating the calculated bubble.

Fig. 9. Experimental

the net local mass efflux J(4) of the gas phase at the
bubble boundary depends only on the angle d, (Fig. 8)
and is given by:
J(4)

= (Epfu.n)

= (~~~u,)sing

Table

+ (~~,u,)cos&
(28)

where n represents the unit outward normal vector

at the bubble boundary. Inspection of Fig. 8 shows
that bubble gas leaks through the bubble roof into
the emulsion phase. However, at the bubble base
emulsion phase gas flows into the bubble. Although
bubble detachment has already occurred at time
t = 0.188 s (corresponding equivalent bubble diameter D, = 0.16 m), substantial inflow of the gas injected
through the orifice still takes place at time f = 0.2 s.
Figure 9 shows the photograph of a detached bubble and the corresponding
density plot at time
t = 0.2 s. The photograph indicates the presence of a
spout between the orifice and the detached gas bubble. By this spout the gas injected through the orifice
is partially introduced into the detached gas bubble.
Obviously, the process of bubble formation is much
more complex than assumed in the approximate
models in the literature. A comparison between the
experimentally observed and the theoretically calculated bubble diameters is presented in Table 5. The
close agreement between the experimental data and
the predictions from the advanced hydrodynamic
model is remarkable. Figure 10 shows a comparison

DV
0,

(photograph)
and theoretical
plot) bubbles.

5. Bubble

diameters

(density

at t = 0.2 s

Experimental

Theoretical

0.174 m
0.164 m
0.170 m

0.172 m
0.182 m
0.164 m

between the observed and calculated bubble shape

factor c = D,/D, during
the process of bubble growth
at the orifice. The deviation from the circular bubble
shape (u = 1.0) appears to be considerable, especially
during the initial stage of bubble formation. For other
orifice velocities than 10 m/s (i.e. 5 and 15 m/s) similar
results have been obtained. Both approximate models
assume circular bubbles during the entire process of
bubble growth.
5.2. Summary ofresults for u, = 5, 10 and 15 mfs
A comparison of the experimental and theoretical
bubble diameters, at the moment of bubble detachment, is presented in Fig. 11 as a function of the orifice
velocity I+,. This figure also shows the predictions
according to the approximate bubble formation models. Both the experimentally determined and the
numerically calculated equivalent bubble diameters
fall between the predictions from these approximate
models. However, in comparing the results it must be

2892
+ Experimental
- Numerical

lj++_
,.,.,.,.
0.00

0.10

0.05

0.20

0.15
t

Fig. 10. Bubble

shape factor

cr during

bubble

__)

data
model

(
0.25

growth at a single orifice in a two-dimensional

bed (u, = 10 m/s).

+ Experimental
n

0.0

5.0

15.0

10.0

u0 +
Fig. 11. Equivalent

bubble

diameter

Numerical

gas-fluid&d

data
model

20.0
(m/s)

D, as a function

Q = ucd;
of the orifice velocity

U,

Table 6. Bubble detachment

time t, and the corresponding (equivalent)
bubble diameter D, as a function of the orifice velocity U,

Experimental

5
10
15

0.160
0.170
0.180

0.096
0.145
0.188

Numerical

5
10
15

0.150
0.188
0.213

0.089
0.160
0.217

5
10
15

0.158
0.200
0.228

0.123
0.195
0.256

5
10
15

Cl.112
0.150
0.178

0.070
0.124
0.171

Approximate
leakage

model with no gas

Approximate
model with leakage of gas at u,, level

realized that the bubble detachment time t, is not the

same in all cases (see Table 6). The numerically calculated bubble detachment time tb falls between the
predictions from the approximate models. Note that

the experimental bubble detachment time is much less

sensitive to the orifice velocity than predicted theoretically from the advanced hydrodynamic model and
the approximate models.

Bubble

formation

at a single orifice in a two-dimensional

6. CONCLUSIONS

J(4)

The formation of gas bubbles at a single orifice in a

two-dimensional
gas-fluidized
bed has been studied
experimentally
and theoretically.
It has been demonstrated that the formation of gas bubbles is predicted
satisfactorily
by the earlier developed first principles
hydrodynamic
model of fluidized beds (Kuipers et al.,
1991). Good agreement between the theoretically predicted bubble sizes and the experimentally
determined
bubble sizes has been obtained.
A comparison
of the present experimental
and
theoretical data with predictions from two approximate models, available in the literature and based on an
idealized picture of the process of bubble formation,
has also been made. Especially during the initial stage
of bubble formation, strong leakage of bubble gas into
the surrounding
porous emulsion phase occurs. This
is indicated by both the results of the experiments and
calculations.
The theoretical
calculations
numerical
additionally
show that during the final stage of bubble
formation, bubble gas leaks through the bubble roof
whereas emulsion phase gas enters the bubble at the
bubble base. The deviatiun from the circular
bubble
shape is considerable
during the initial stage of bubble
formation.
Although
the approximate
models clearly have
their utility because of their simplicity and ability to
oredict the correct order of magnitude of the bubble
&es, they do not give a correct representation
of these
phenomena.
This investigation
was supported by
B. V. of the Netherlands. central technical
institute
of the Dutch gas supply companies. We also
acknowledge our indebtedness
to A. R. Roebersen and
H. Tammes for their assistance in the experimental
work and
Acknowledgements-

VEG-Gasinstituut

the numerical

solution

of the approximate

theoretical

mo-

dels.
NOTATION

D,

bubble surface, m2
virtual mass coefficient
drag coefficient
compaction
modulus
bubble diameter, m
eouivalent bubble diameter,

horizontal

D,

vertical bubble diameter, m

bubble diameter for the case of zero gas leakage, m
bed diameter, m
orifice diameter, m
particle diameter, m
interaction
function defined in eq. (18b)
particle-particle
interaction
modulus, Pa
particle-particle
interaction
modulus
for
E = E*, Pa
gravitational
force per unit mass, m/s
bed height at minimum fluidization conditions,
m
unit tensor

A,
CO
Cl
C
D,

Do

L,

bubble

diameter,

M
n

nb

Q
R

Rb

Rep
rb

v,
VCI
V
X

gas-fluid&d

bed

2893

mass flux of the gas phase at the bubble boundary defined in eq. (28), kg/(m* s)
molecular weight, kg/kmol
unit outward normal vector at bubble boundary
bubble frequency, s - 1
pressure, Pa
gas flow rate through
orifice, m3js
gas constant,
J/(kmol K)
bubble radius, m
particle Reynolds number
equivalent bubble radius, m
bubble centre
temperature,
K
time, s
bubble formation time, s
fluid phase velocity, m/s
minimum fluidization velocity, m/s
superficial
injection
velocity through

orifice,

m/s
bubble volume, m3
bubble volume for the case of zero gas leakage,
m3
solid phase velocity, m/s
lateral coordinate,
m
vertical coordinate,
m

Greek
letters
_
volumetric interphase momentum transfer coefcr
ficient, kg/(m s)
bed thickness, m
lateral computational
cell dimension, m
6x
vertical
computational
cell dimension, m
6Y
E
porosity
minimum fluidization porosity
M
e*
compaction
gas phase volume fraction
shear viscosity, kg/(m s)
cc
density, kg/m3
P
bubble shape factor defined in Fig. 10
angle defined in Fig. 8
;
sphericity
6%
integral leakage fraction defined in eq. (27)
r(t)
integral gas leakage during process of bubble
n

formation,

m3

Subscripts
b
bed
bubble
equivalent
e
emulsion phase
fluid phase
f
h
horizontal
max maximum
mf minimum fluidization
microscopic
property
0
particle
P
solid phase
vertical
V
x-direction
X
y-direction
freeboard conditons
0

conditions

2894
Superscripts
T
transpose
*
asymptotic

J. A. M.

KUIPERS

value

Operator
V
gradient
Vdivergence

REFERENCES
Caram, H. S. and Hsu, K. K., 1986, Chem. Engng Sci. 41,
1445.
Davidson,
J. F. and Schiiler, B. 0. G., 1960, Trans. lnstn
them. Engrs 38, 335.
Davidson, J. F. and Harrison, D., 1963, Fluidized Particks.
Cambridge University Press, Cambridge
Grace, J. R., 1970, Can. J. them. Engng 48, 30.
Harrison, D. and Leung, L. S., 1961, Trans. lnstn ckem. Engrs
39, 409.
Kuipers, J. A. M., van Duin, K. J., van B&urn, F. P. H. and
van Swaaij, W. P. M., 1991, A numerical model of gas-

el al.

fluidized beds. Chem. Engng Sci. (submitted).

Milne-Thomson,
L. M., 1960, Theoretical Hydrodynamics,
4th Edition. Macmillan, London.
Nguyen, X. T. and Leung, L. S., 1972, Chem. Engng Sci. 27,
I TAR
_,
._.
Radestock.
J. and Jeschar, R., 1971, Chemir-lnyr-Tech. 43,
1304.
Rowe, P. N. and Henwood, G. A., 1961, Trans. lnstn rhem.
Engrs 39, 43.
Rowe, P. N., 1961, Trans. Instn them. Engrs 39, 175.
Rowe, P. N., 1971, in Fluidiration (Edited by J. F. Davidson,
and D. Harrison), p. 121. Academic Press, New York.
Rowe, P. N., MacGillivray,
H. J. and Cheesman, D. J., 1979,
Trans. Instn them. Engrs 57, 194.
Schiigerl, K., Merz, M. and Fetting, F., 1961, Chem. Engng
sci. 15, 1.
Wen, Y. C. and Yu, Y. H., 1966, Chem. Engng Prog. Symp.
Ser. 62, 100.
Yang, W.-C., Revay, D., Anderson,
R. G., Chelen, E. J.,
Keairns, D. L. and Cicero, D. C., 1984, in FIuidization
(Edited by D. Kunii, and R. Toei), p. 77. Engineering
Foundation,
distributed
by A.1.Ch.E. J.
Zenz, F. A., 1968, Inst. them. Engng Symp. Ser. 30, 136.